Distance metric learning using random forest for cytometry data.

Visualization and clustering of single-cell mass cytometry (CyTOF) data are analytic techniques to identify different cell types. Most of such techniques, such as Euclidean norm, lose their effectiveness when the data dimension increases due to the curse of dimensionality. In this paper, we propose a new cell distance (called CytoRFD) that works based on Random Forest (RF) concept. The experimental results show that the proposed distance can achieve a much higher quality and effectiveness in large data analysis than traditional metrics specially for CyTOF data.

An adaptive deep learning approach for PPG-based identification.

Wearable biosensors have become increasingly popular in healthcare due to their capabilities for low cost and long term biosignal monitoring. This paper presents a novel two-stage technique to offer biometric identification using these biosensors through Deep Belief Networks and Restricted Boltzman Machines. Our identification approach improves robustness in current monitoring procedures within clinical, e-health and fitness environments using Photoplethysmography (PPG) signals through deep learning classification models. The approach is tested on TROIKA dataset using 10-fold cross validation and achieved an accuracy of 96.1%.

Clustering-based limb identification for pressure ulcer risk assessment.

Bedridden patients have a high risk of developing pressure ulcers. Risk assessment for pressure ulceration is critical for preventive care. For a reliable assessment, we need to identify and track the limbs continuously and accurately. In this paper, we propose a method to identify body limbs using a pressure mat. Three prevalent sleep postures (supine, left and right postures) are considered. Then, predefined number of limbs (body parts) are identified by applying Fuzzy C-Means (FCM) clustering on key attributes. We collected data from 10 adult subjects and achieved average accuracy of 93.2% for 10 limbs in supine and 7 limbs in left/right postures.

Sleep state classification using pressure sensor mats.

Sleep state detection is valuable in assessing patient's sleep quality and in-bed general behavior. In this paper, a novel classification approach of sleep states (sleep, pre-wake, wake) is proposed that uses only surface pressure sensors. In our method, a mobility metric is defined based on successive pressure body maps. Then, suitable statistical features are computed based on the mobility metric. Finally, a customized random forest classifier is employed to identify various classes including a new class for pre-wake state. Our algorithm achieves 96.1% and 88% accuracies for two (sleep, wake) and three (sleep, pre-wake, wake) class identification, respectively.

Classifying bed inclination using pressure images.

Pressure ulcer is one of the most prevalent problems for bed-bound patients in hospitals and nursing homes. Pressure ulcers are painful for patients and costly for healthcare systems. Accurate in-bed posture analysis can significantly help in preventing pressure ulcers. Specifically, bed inclination (back angle) is a factor contributing to pressure ulcer development. In this paper, an efficient methodology is proposed to classify bed inclination. Our approach uses pressure values collected from a commercial pressure mat system. Then, by applying a number of image processing and machine learning techniques, the approximate degree of bed is estimated and classified. The proposed algorithm was tested on 15 subjects with various sizes and weights. The experimental results indicate that our method predicts bed inclination in three classes with 80.3% average accuracy.

A wrist-worn biosensor system for assessment of neurological status.

EEG based monitoring for the purpose of assessing a patient's neurological status is conspicuous and uncomfortable at best. We are analyzing a set of physiological signals that may be monitored comfortably by a wrist worn device. We have found that these signals and machine based classification allows us to accurately discriminate among four stress states of individuals. Further, we have found a clear change in these signals during the 70 minutes preceding a single convulsive epileptic seizure. Our classification accuracy on all data has been greater than 90% to date.

In vitro study of antibacterial activity of the alga Sargassum oligocystum from the Persian Gulf.

BACKGROUND AND OBJECTIVES: With due attention to the development of drug-resistant bacteria, discovering of new antibacterial compounds is needed. Algae produce numerous bioactive substances which may have pharmacological properties such as antibacterial activity. The objective of this investigation was to in vitro study of antibacterial activity of brown alga Sargassum oligocystum collected along the Bushehr coast of Persian Gulf (south west of Iran). MATERIALS AND METHODS: Hot water extract, cold water extract, and hot glycerin extract were prepared. The effect of the extracts were investigated on Staphylococcus aureus (ATCC 25923), Staphylococcus epidermidis (ATCC 14990), Pseudomonas aeruginosa (ATCC 27853), and Escherichia coli (ATCC 25922). RESULTS: Hot water extract exhibited antibacterial activity against Staphylococcus aureus, Staphylococcus epidermidis, and Pseudomonas aeruginosa. Cold water extract and hot glycerin extract did not show antibacterial activity on any of the four test bacteria. The minimum inhibitory concentration (MIC) of hot water extract for both Staphylococcus aureus and Staphylococcus epidermidis was 3.175 mg/ml. However, the MIC of this extract for Pseudomonas aeruginosa was 9.556 mg/ml. DISCUSSION: In this study gram-positive bacteria were more susceptible to hot water extract than gram-negative bacteria. Extract of Sargassum oligocystum could be a candidate for purification and further in vivo studies.